marketing-cro
Compare original and translation side by side
🇺🇸
Original
English🇨🇳
Translation
ChineseCRO — CONVERSION OPTIMIZATION OS (OPERATIONAL)
CRO — 转化率优化运营体系（实操向）
Built as a no-fluff execution skill for systematic conversion rate optimization.
Structure: Core CRO fundamentals first. Advanced testing in dedicated sections. AI/ML optimization in clearly labeled "Optional: AI / Automation" sections.
这是一套专为系统化转化率优化打造的干货型实操技能指南。
结构：先讲解CRO核心基础，再在专属章节介绍进阶测试内容，AI/ML优化部分则标注为“可选：AI/自动化”板块。
Modern Best Practices (January 2026)
2026年1月现代最佳实践
- Google Optimize sunset: Use VWO, Optimizely, or PostHog
- Statistical significance: https://www.evanmiller.org/ab-testing/
- CXL Institute: https://cxl.com/
- Baymard Institute UX: https://baymard.com/
- Cookie deprecation + stricter privacy defaults: prefer first-party measurement, validate assignment/tracking, and treat lifts as uncertain without clean instrumentation
- Google Optimize已停用：建议使用VWO、Optimizely或PostHog
- 统计显著性工具：https://www.evanmiller.org/ab-testing/
- CXL Institute：https://cxl.com/
- Baymard Institute UX：https://baymard.com/
- Cookie废弃与更严格的隐私默认设置：优先使用第一方数据统计，验证分配/追踪机制；若无完善的监测工具，请勿轻信转化提升数据
When to Use This Skill
适用场景
- Landing page optimization: Hero, CTA, proof, form optimization
- A/B testing: Hypothesis design, sample size, statistical significance
- Funnel analysis: Drop-off identification, micro-conversion mapping
- Form optimization: Field reduction, multi-step forms, friction removal
- Trust/credibility: Social proof, security signals, guarantees
- 着陆页优化：首屏内容、CTA、信任背书、表单优化
- A/B测试：假设设计、样本量计算、统计显著性验证
- 漏斗分析：流失节点识别、微转化映射
- 表单优化：字段精简、多步骤表单、减少填写阻力
- 信任/可信度提升：社交证明、安全标识、售后保障
When NOT to Use
不适用场景
- Brand awareness campaigns → Use marketing-paid-advertising
- User research methodology → Use software-ux-research
- Product analytics setup → Use marketing-product-analytics
- SEO/organic traffic → Use marketing-seo-complete
- 品牌知名度推广 → 参考marketing-paid-advertising
- 用户研究方法论 → 参考software-ux-research
- 产品分析搭建 → 参考marketing-product-analytics
- SEO/自然流量 → 参考marketing-seo-complete
Expert: CRO Mental Model (Quick Calibration)
专家视角：CRO思维模型（快速校准）
Use this to avoid local wins / global losses.
- CRO: Increase the rate of valuable commitments (purchase, qualified lead, activation) while protecting business outcomes (revenue, margin, LTV, support load).
- UX optimization: Reduce friction/errors so users can do what they already intend; good UX does not guarantee better conversions.
- Funnel optimization: Optimize the system across steps and handoffs (traffic quality → intent match → page → form/checkout → sales/onboarding → retention).
- Experimentation: A causal learning method; not every decision belongs in a test.
Do not delegate these to A/B tests (even with infinite traffic): legal/compliance/ethics, dark patterns, misleading claims, and irreversible brand trust decisions.
用于避免局部优化但全局受损的情况。
- CRO：在保障业务成果（收入、利润率、客户终身价值、支持成本）的前提下，提升有价值行为的转化率（购买、合格线索、激活）。
- UX优化：减少操作阻力/错误，让用户顺利完成原本就想做的事；优质UX不代表更高转化率。
- 漏斗优化：跨步骤和环节优化整个体系（流量质量→意图匹配→页面→表单/结账→销售/入职→留存）。
- 实验测试：一种因果学习方法；并非所有决策都需要测试。
以下内容不要交给A/B测试（即使流量充足）：法律/合规/伦理问题、诱导性设计、误导性声明、不可逆的品牌信任决策。
Core: CRO Framework
核心：CRO框架
The CRO Process
CRO流程
text
1. ANALYZE → Identify conversion problems (data + qualitative)
2. HYPOTHESIZE → Form testable hypotheses
3. PRIORITIZE → Score by impact/effort (ICE/PIE)
4. TEST → Run A/B tests with statistical rigor
5. LEARN → Document results, iterate
6. IMPLEMENT → Roll out winners, test nexttext
1. 分析 → 识别转化问题（数据+定性研究）
2. 假设 → 形成可测试的假设
3. 优先级排序 → 按影响/成本评分（ICE/PIE框架）
4. 测试 → 严格按照统计要求开展A/B测试
5. 学习 → 记录结果，迭代优化
6. 落地 → 推广测试获胜方案，开展下一轮测试Conversion Rate Benchmarks
转化率基准
| Page Type | Poor | Average | Good | Great |
|---|---|---|---|---|
| Landing page | <1% | 2-3% | 4-5% | >6% |
| Checkout | <40% | 50-60% | 65-75% | >80% |
| Form completion | <20% | 30-40% | 45-55% | >60% |
| Add to cart | <3% | 5-8% | 9-12% | >15% |
Note: Benchmarks vary significantly by industry. Use as directional only.
| 页面类型 | 较差 | 一般 | 良好 | 优秀 |
|---|---|---|---|---|
| 着陆页 | <1% | 2-3% | 4-5% | >6% |
| 结账页 | <40% | 50-60% | 65-75% | >80% |
| 表单完成率 | <20% | 30-40% | 45-55% | >60% |
| 加购率 | <3% | 5-8% | 9-12% | >15% |
注：基准数据因行业差异较大，仅作方向参考。
Core: Landing Page Optimization
核心：着陆页优化
Above-the-Fold Checklist
首屏检查清单
Every landing page needs these elements visible without scrolling:
| Element | Requirement | Common Issues |
|---|---|---|
| Headline | Clear value proposition | Vague, company-focused |
| Subheadline | Specific benefit or outcome | Missing or weak |
| Hero image/video | Relevant, shows outcome | Stock photos, irrelevant |
| CTA | Prominent, action-oriented | Hidden, generic text |
| Trust signal | Logo strip, rating, or stat | Missing entirely |
所有着陆页的首屏（无需滚动可见区域）必须包含以下元素：
| 元素 | 要求 | 常见问题 |
|---|---|---|
| 标题 | 清晰传达价值主张 | 表述模糊、以企业为中心 |
| 副标题 | 明确说明具体收益或成果 | 缺失或内容薄弱 |
| 首屏图片/视频 | 内容相关，展示成果 | 使用通用图库、内容无关 |
| CTA按钮 | 突出显示、行动导向 | 隐藏、文字通用化 |
| 信任标识 | 品牌logo墙、评分或数据 | 完全缺失 |
Headline Formula
标题公式
text
[Outcome] + [Timeframe/Ease] + [Without Pain Point]
Examples:
"Get 10 qualified leads per week without cold calling"
"File your tax return in 15 minutes with expert review"
"Double your email conversions without hiring a copywriter"text
[成果] + [时间/便捷性] + [无痛点]
示例：
"无需陌生电话开发，每周获取10条合格线索"
"15分钟完成报税，还有专家审核"
"无需雇佣文案，邮件转化率翻倍"CTA Button Best Practices
CTA按钮最佳实践
| Do | Don't |
|---|---|
| "Start Free Trial" | "Submit" |
| "Get My Quote" | "Click Here" |
| "Book My Demo" | "Learn More" (bottom of funnel) |
| "Download the Guide" | "Send" |
CTA Button Optimization:
- Size: Large enough to tap on mobile (min 44px height)
- Color: Contrasts with page background
- Position: Above fold AND after key sections
- Text: First person ("Get My...") often outperforms second person
- Whitespace: Use spacing to isolate the primary CTA from competing elements; treat big lift claims as case-dependent and verify in your context
| 推荐 | 避免 |
|---|---|
| "开始免费试用" | "提交" |
| "获取我的报价" | "点击这里" |
| "预约演示" | "了解更多"（漏斗底部场景） |
| "下载指南" | "发送" |
CTA按钮优化要点：
- 尺寸：移动端点击高度至少44px
- 颜色：与页面背景形成对比
- 位置：首屏及关键内容板块后均需设置
- 文字：第一人称表述（"获取我的..."）通常优于第二人称
- 留白：用空白区域突出主CTA，避免与其他元素竞争；转化提升效果需结合自身场景验证，勿轻信通用结论
Trust Elements Hierarchy
信任元素层级
text
STRONGEST TRUST SIGNALS (use at least 3):
├─ Customer logos (recognizable brands)
├─ Review score (4.5+ stars with count)
├─ Security badges (SSL, payment, compliance)
├─ Money-back guarantee
└─ Phone number visible
SUPPORTING TRUST SIGNALS:
├─ Customer testimonials (with photo, name, company)
├─ Case study snippets (specific metrics)
├─ "As seen in" media logos
├─ Team photos (for services)
├─ Live chat widget
└─ Physical address (for services)text
最强信任标识（至少使用3种）：
├─ 客户logo（知名品牌）
├─ 评分（4.5星以上，带评价数量）
├─ 安全徽章（SSL、支付、合规认证）
├─ 退款保障
└─ 可见的联系电话
辅助信任标识：
├─ 客户 testimonial（带照片、姓名、公司）
├─ 案例研究片段（含具体数据）
├─ "被以下媒体报道" logo墙
├─ 团队照片（服务类产品）
├─ 在线聊天组件
└─ 实体地址（服务类产品）User-Generated Content (UGC)
用户生成内容（UGC）
UGC often increases conversions in SaaS and e-commerce, but lift magnitude varies widely by category, placement, and traffic intent.
| UGC Type | Placement | Impact |
|---|---|---|
| Customer videos | Hero or below fold | High trust, high engagement |
| Review excerpts | Near CTA | Reduces uncertainty |
| Case study quotes | Consideration section | Builds credibility |
| Community mentions | Footer or social proof bar | Volume signal |
Implementation: Pull from G2, Capterra, or in-app feedback. Verify permissions before use.
UGC通常能提升SaaS和电商的转化率，但提升幅度因品类、放置位置和流量意图差异较大。
| UGC类型 | 放置位置 | 影响 |
|---|---|---|
| 客户视频 | 首屏或首屏下方 | 高信任度、高参与度 |
| 评价摘录 | CTA附近 | 降低用户顾虑 |
| 案例研究引用 | 决策考量板块 | 提升可信度 |
| 社区提及 | 页脚或社交证明栏 | 体现用户规模 |
实施建议：从G2、Capterra或应用内反馈获取内容，使用前需确认权限。
Core: Form Optimization
核心：表单优化
Form Field Rules
表单字段规则
| Rule | Why | Impact |
|---|---|---|
| Minimum fields | Every field adds friction | Often lowers completion (magnitude varies) |
| Email first | Captures partial submissions | +15-30% lead capture |
| Persistent labels | Placeholders disappear, cause errors | +10% completion |
| Single column | Easier flow | +5-10% completion |
| Inline validation | Catch errors early | +22% completion |
| Browser autofill | Reduces typing, fewer errors | +15-20% completion |
2026 Benchmark: Average checkout = 5.1 steps, 11.3 fields (Baymard). Target ≤5 fields for lead gen.
| 规则 | 原因 | 影响 |
|---|---|---|
| 最少字段 | 每增加一个字段就会增加填写阻力 | 通常会降低完成率（幅度因场景而异） |
| 先收集邮箱 | 可捕获部分提交数据 | 线索捕获量提升15-30% |
| 固定标签 | 占位符会消失，易导致错误 | 完成率提升10% |
| 单列布局 | 流程更顺畅 | 完成率提升5-10% |
| 实时验证 | 提前发现错误 | 完成率提升22% |
| 浏览器自动填充 | 减少输入，降低错误 | 完成率提升15-20% |
2026年基准：平均结账流程为5.1步骤、11.3个字段（Baymard数据）。线索收集表单目标字段数≤5个。
Field Priority (Ask Only What You Need)
字段优先级（仅收集必要信息）
| Priority | Field | When Required |
|---|---|---|
| 1 | Always | |
| 2 | Name | If personalization needed |
| 3 | Company | B2B only |
| 4 | Phone | Sales-ready leads only |
| 5 | Job title | Enterprise targeting |
| 6+ | Everything else | Gate behind progressive profiling |
| 优先级 | 字段 | 收集场景 |
|---|---|---|
| 1 | 邮箱 | 始终需要 |
| 2 | 姓名 | 需要个性化时 |
| 3 | 公司 | 仅B2B场景 |
| 4 | 电话 | 仅针对可跟进的成熟线索 |
| 5 | 职位 | 仅针对企业级客户 |
| 6+ | 其他所有字段 | 通过渐进式收集逐步获取 |
Multi-Step Form Pattern
多步骤表单模式
text
Step 1: Low commitment (email)
├─ "What's your email?"
├─ Progress indicator: 1 of 3
└─ CTA: "Continue"
Step 2: Qualifying info
├─ Company size / Industry
├─ Progress indicator: 2 of 3
└─ CTA: "Almost there"
Step 3: Contact info
├─ Name / Phone (optional)
├─ Progress indicator: 3 of 3
└─ CTA: "Get My [Deliverable]"Multi-step benefits:
- Commitment and consistency principle
- Captures partial data (even if abandoned)
- Feels less overwhelming
- Can qualify leads progressively
text
步骤1：低承诺（邮箱）
├─ "请输入您的邮箱？"
├─ 进度提示：3步中的第1步
└─ CTA："继续"
步骤2：资格验证信息
├─ 公司规模 / 行业
├─ 进度提示：3步中的第2步
└─ CTA："即将完成"
步骤3：联系信息
├─ 姓名 / 电话（可选）
├─ 进度提示：3步中的第3步
└─ CTA："获取我的[资料]"多步骤表单优势：
- 利用承诺与一致原则
- 即使用户中途放弃也能捕获部分数据
- 感觉更轻松，无压迫感
- 可逐步筛选合格线索
Core: A/B Testing Methodology
核心：A/B测试方法论
Hypothesis Template
假设模板
text
IF we [change/add/remove X]
THEN [metric] will [increase/decrease] by [estimate]
BECAUSE [reasoning based on data/research]
Example:
IF we add customer logos to the hero section
THEN form conversion will increase by 15%
BECAUSE trust signals reduce perceived risk for new visitorstext
如果我们[添加/修改/移除X]
那么[指标]将[提升/下降][预估幅度]
因为[基于数据/研究的理由]
示例：
如果我们在首屏添加客户logo墙
那么表单转化率将提升15%
因为信任标识能降低新访客的感知风险Sample Size Calculator
样本量计算器
Minimum sample size formula (simplified):
text
n = (16 × p × (1-p)) / MDE²
Where:
- n = sample per variant
- p = baseline conversion rate
- MDE = minimum detectable effect (e.g., 0.10 for 10% lift)
Example:
Baseline CVR: 3% (0.03)
MDE: 20% relative lift (looking for 3.6% or higher)
n = (16 × 0.03 × 0.97) / (0.006)²
n ≈ 12,933 per variant
Total traffic needed: ~26,000 visitorsQuick reference:
| Baseline CVR | 10% MDE | 20% MDE | 30% MDE |
|---|---|---|---|
| 1% | 63,000 | 15,800 | 7,000 |
| 3% | 20,700 | 5,200 | 2,300 |
| 5% | 12,200 | 3,050 | 1,350 |
| 10% | 5,800 | 1,450 | 650 |
Per variant. Multiply by 2 for total traffic needed.
简化版最小样本量公式：
text
n = (16 × p × (1-p)) / MDE²
参数说明：
- n = 每个变体的样本量
- p = 基准转化率
- MDE = 最小可检测效果（如10%提升则为0.10）
示例：
基准转化率：3%（0.03）
MDE：20%相对提升（目标转化率3.6%及以上）
n = (16 × 0.03 × 0.97) / (0.006)²
n ≈ 12933（每个变体）
所需总流量：约26000访客快速参考表：
| 基准转化率 | 10% MDE | 20% MDE | 30% MDE |
|---|---|---|---|
| 1% | 63000 | 15800 | 7000 |
| 3% | 20700 | 5200 | 2300 |
| 5% | 12200 | 3050 | 1350 |
| 10% | 5800 | 1450 | 650 |
每个变体的样本量。总流量需乘以2。
Statistical Significance
统计显著性
Requirements for valid test:
- 95% confidence level (minimum)
- 80% power (default) unless you have a reason to change it
- Run for at least 1-2 full business cycles (7-14 days)
- Don't peek and stop early (increases false positives)
- Document before test: hypothesis, primary metric, guardrails, sample size, duration
- Avoid post-hoc slicing; pre-register segments or adjust for multiple comparisons
Reality check (expert defaults):
- Statistical significance does not mean the change is worth shipping (check practical impact + guardrails)
- Ignore "significant" results when experiment integrity is in doubt (tracking issues, traffic mix shifts, SRM, broken randomization)
- Stop early only for clear harm (guardrail breaches) or invalidity (instrumentation/assignment problems), not for "early wins"
有效测试要求：
- 至少95%置信水平
- 默认80%统计功效（除非有特殊理由调整）
- 至少运行1-2个完整业务周期（7-14天）
- 不要中途偷看数据并提前停止（会增加假阳性概率）
- 测试前记录：假设、核心指标、防护指标、样本量、时长
- 避免事后细分分析；提前注册细分群体或调整多重比较的统计阈值
专家默认准则：
- 统计显著性不代表该改动值得上线（需结合实际影响+防护指标判断）
- 若实验完整性存疑（追踪问题、流量结构变化、SRM、随机分组失效），忽略“显著”结果
- 仅在出现明显损害（防护指标触发）或实验无效（监测/分组问题）时提前停止，不要因“早期获胜”提前结束
Experiment Integrity (2026 Default Checks)
实验完整性检查（2026年默认项）
- Assignment sanity: A/A test periodically; check SRM on day 1 and day 3
- Tracking sanity: confirm event definitions, dedupe, cross-domain, and consent-mode behavior before interpreting results
- Contamination: avoid showing multiple variants to the same user across devices/sessions; prefer stable IDs when possible
- Change control: freeze other major changes to the same flow during the test window
- 分组合理性：定期开展A/A测试；在测试第1天和第3天检查SRM（样本比例不匹配）
- 追踪合理性：在解读结果前确认事件定义、去重、跨域、 consent-mode行为
- 污染防控：避免同一用户在不同设备/会话中看到多个变体；尽可能使用稳定用户ID
- 变更控制：测试期间冻结对同一流程的其他重大变更
CUPED: Faster Tests via Variance Reduction
CUPED：通过方差缩减加速测试
CUPED (Controlled-experiment Using Pre-Existing Data) can reduce variance by ~40-60%, allowing tests to reach significance faster.
| Aspect | Details |
|---|---|
| How it works | Uses pre-experiment user behavior to control for inherent variance |
| Lookback window | 1-2 weeks (optimal balance) |
| Limitation | Doesn't work for new users (no history) |
| Platforms | VWO, Optimizely, Statsig, Eppo, PostHog |
When to use: High-traffic sites where test velocity matters. See advanced-testing.md for implementation details.
CUPED（利用历史数据的受控实验）可将方差降低约40-60%，让测试更快达到统计显著性。
| 维度 | 详情 |
|---|---|
| 工作原理 | 利用实验前的用户行为数据控制固有方差 |
| 回溯窗口 | 1-2周（最优平衡） |
| 局限性 | 对新用户无效（无历史数据） |
| 支持平台 | VWO、Optimizely、Statsig、Eppo、PostHog |
适用场景：高流量网站，测试速度至关重要。实现细节请参考advanced-testing.md。
Test Prioritization: ICE Framework
测试优先级：ICE框架
| Factor | Score (1-10) | Description |
|---|---|---|
| Impact | How much will this move the metric? | |
| Confidence | How sure are we this will work? | |
| Ease | How easy is this to implement? | |
| ICE Score | (Impact + Confidence + Ease) / 3 |
ICE Score interpretation:
- 8-10: High priority, test immediately
- 5-7: Medium priority, add to queue
- 1-4: Low priority, revisit later or skip
| 因素 | 评分（1-10） | 说明 |
|---|---|---|
| 影响 | 对指标的提升幅度有多大？ | |
| 信心 | 对该改动有效的把握有多大？ | |
| 成本 | 实现难度有多低？ | |
| ICE得分 | （影响+信心+成本）/3 |
ICE得分解读：
- 8-10：高优先级，立即测试
- 5-7：中优先级，加入测试队列
- 1-4：低优先级，后续再评估或跳过
Core: Funnel Analysis
核心：漏斗分析
Funnel Diagnostic Framework
漏斗诊断框架
text
STEP 1: Map your funnel
Page Visit → Key Action → Form Start → Form Complete → Confirmation
STEP 2: Measure drop-off at each step
├─ Page Visit to Key Action: ___% (bounce rate inverse)
├─ Key Action to Form Start: ___%
├─ Form Start to Complete: ___%
└─ Complete to Confirmation: ___%
STEP 3: Identify biggest drop-off
Biggest percentage drop = highest priority to fix
STEP 4: Diagnose root cause
├─ High bounce? → Relevance, load speed, messaging
├─ Low engagement? → Content, CTA visibility
├─ Form abandonment? → Form friction, trust
└─ Checkout drop? → Pricing, shipping, trustExpert note: The "biggest drop-off" is not always the best target. Confirm it's a defect (not intentional filtering), not a measurement artifact, and not caused upstream (traffic quality / offer mismatch).
text
步骤1：绘制漏斗图
页面访问 → 关键行动 → 开始填写表单 → 完成表单 → 确认
步骤2：测量各步骤流失率
├─ 页面访问到关键行动：___%（跳出率的倒数）
├─ 关键行动到开始填写表单：___%
├─ 开始填写到完成表单：___%
└─ 完成表单到确认：___%
步骤3：识别最大流失节点
流失率最高的步骤 = 优先优化目标
步骤4：诊断根本原因
├─ 高跳出率？→ 相关性、加载速度、信息传达
├─ 低参与度？→ 内容、CTA可见性
├─ 表单放弃？→ 表单阻力、信任问题
└─ 结账流失？→ 价格、运费、信任问题专家提示：“最大流失节点”不一定是最佳优化目标。需确认这是真实问题（而非有意筛选）、不是统计 artifact、也不是上游因素导致（流量质量/offer不匹配）。
Micro-Conversion Mapping
微转化映射
| Funnel Stage | Micro-Conversions to Track |
|---|---|
| Awareness | Scroll depth, time on page, video views |
| Interest | CTA hover, tab/section views, resource clicks |
| Consideration | Pricing page visit, comparison page, demo video |
| Decision | Form start, add to cart, checkout start |
| Conversion | Form complete, purchase, signup |
| 漏斗阶段 | 需追踪的微转化 |
|---|---|
| 认知阶段 | 滚动深度、页面停留时间、视频观看量 |
| 兴趣阶段 | CTA悬停、板块浏览、资源点击 |
| 考虑阶段 | 定价页访问、对比页访问、演示视频观看 |
| 决策阶段 | 开始填写表单、加购、开始结账 |
| 转化阶段 | 完成表单、购买、注册 |
Heatmap & Recording Analysis
热力图与会话录屏分析
What to look for:
- Click heatmaps: Are users clicking CTAs? Clicking non-clickable elements?
- Scroll maps: Where do users stop scrolling? Key content below fold?
- Session recordings: Where do users hesitate? Rage clicks? Form confusion?
- Form analytics: Which fields cause abandonment? Error patterns?
关注要点：
- 点击热力图：用户是否点击CTA？是否点击不可点击元素？
- 滚动热力图：用户在哪里停止滚动？关键内容是否在首屏以下？
- 会话录屏：用户在哪里犹豫？是否有愤怒点击？是否对表单感到困惑？
- 表单分析：哪些字段导致放弃？错误模式是什么？
Reference: Triage, Speed, SOPs
参考：分流、速度、SOP
For page speed targets, CRO triage decision tree, operating cadence, and anti-patterns, see .
references/triage-and-ops.md页面速度目标、CRO分流决策树、运营节奏及反模式，请参考。
references/triage-and-ops.mdTemplates
模板
| Template | Purpose |
|---|---|
| landing-audit.md | Full landing page audit |
| ab-test-plan.md | A/B test planning |
| form-audit.md | Form optimization checklist |
| funnel-analysis.md | Funnel diagnostic |
| ice-scoring.md | Test prioritization |
| 模板 | 用途 |
|---|---|
| landing-audit.md | 完整着陆页审核 |
| ab-test-plan.md | A/B测试规划 |
| form-audit.md | 表单优化检查清单 |
| funnel-analysis.md | 漏斗诊断 |
| ice-scoring.md | 测试优先级排序 |
Expert: Hypothesis Quality (Silent Failure Checklist)
专家：假设质量（隐性失败检查清单）
A good CRO hypothesis is not "change X to raise CVR." It must specify mechanism and risk.
Strong hypothesis includes:
- Which constraint it targets: clarity, trust, motivation, friction
- Who it's for: segment/intent/channel/device (at least one)
- What moves: primary metric + guardrails (value, quality, downstream)
- Why it should work: evidence + mechanism (not vibes)
How CRO fails silently (common):
- Conversions go up but value goes down (lower-quality leads, higher refunds/chargebacks, worse retention)
- Overall looks flat but a high-value segment is harmed (mix effects hide damage)
- "Win" is novelty or seasonality; it doesn't repeat
Use to pre-register guardrails and invalidation criteria.
assets/ab-test-plan.md优秀的CRO假设不是“修改X以提升转化率”，必须明确机制和风险。
优质假设包含：
- 针对的约束：清晰度、信任、动机、阻力
- 目标用户：细分群体/意图/渠道/设备（至少一项）
- 影响指标：核心指标+防护指标（价值、质量、下游影响）
- 理由：基于证据+机制（而非主观感受）
CRO隐性失败的常见原因：
- 转化率提升但用户价值下降（低质量线索、高退款/拒付、留存变差）
- 整体数据持平但高价值群体受损（混合效应掩盖损害）
- “获胜”是因为新鲜感或季节性因素，无法复制
使用提前注册防护指标和无效判定标准。
assets/ab-test-plan.mdReferences
参考资料
| Reference | Description |
|---|---|
| advanced-testing.md | CUPED, sequential testing, MAB |
| ai-automation.md | AI personalization, tool stack |
| triage-and-ops.md | Page speed, triage, SOPs, anti-patterns |
| 参考 | 说明 |
|---|---|
| advanced-testing.md | CUPED、序贯测试、多臂老虎机 |
| ai-automation.md | AI个性化、工具栈 |
| triage-and-ops.md | 页面速度、分流、SOP、反模式 |
International Markets
国际市场
This skill uses US/UK defaults. For international CRO:
| Need | See Skill |
|---|---|
| Regional payment methods | marketing-geo-localization |
| Cultural trust signals | marketing-geo-localization |
| Regional CTA adaptation | marketing-geo-localization |
| RTL/localized design | marketing-geo-localization |
Auto-triggers: When your query mentions regional markets or cultural adaptation, both skills load automatically.
本指南默认采用美/英地区规则。针对国际市场CRO：
| 需求 | 参考技能 |
|---|---|
| 区域支付方式 | marketing-geo-localization |
| 文化适配信任标识 | marketing-geo-localization |
| 区域CTA适配 | marketing-geo-localization |
| RTL/本地化设计 | marketing-geo-localization |
自动触发：当查询提及区域市场或文化适配时，将自动加载本技能及上述技能。
Related Skills
相关技能
- marketing-geo-localization — International markets, cultural CRO
- marketing-leads-generation — Lead capture strategies
- marketing-paid-advertising — Traffic sources
- marketing-seo-complete — Page speed, Core Web Vitals
- software-ui-ux-design — Design patterns
- software-ux-research — User research methods
- marketing-geo-localization — 国际市场、文化适配CRO
- marketing-leads-generation — 线索捕获策略
- marketing-paid-advertising — 流量来源
- marketing-seo-complete — 页面速度、Core Web Vitals
- software-ui-ux-design — 设计模式
- software-ux-research — 用户研究方法
Usage Notes (Claude)
使用说明（Claude）
- Stay operational: return checklists, audit results, test plans
- Always include statistical significance requirements for testing
- Recommend qualitative research for low-traffic sites
- Use benchmark ranges, not absolute numbers
- Do not invent conversion data; state "varies by industry" when uncertain
- 聚焦实操：返回检查清单、审核结果、测试计划
- 测试相关内容必须包含统计显著性要求
- 低流量网站建议结合定性研究
- 使用基准范围，而非绝对数值
- 不要编造转化数据；不确定时注明“因行业而异”",